molecular modelling and computer-aided drug design: the
TRANSCRIPT
Hacettepe University Journal of the Faculty of PharmacyVolume 40 / Number 1 / January 2020 / pp. 34-47
Molecular Modelling and Computer-Aided Drug Design: The Skill Set Every Scientist in Drug Research Needs and Can Easily GetMoleküler Modelleme ve Bilgisayar Destekli İlaç Tasarımı: İlaç Araştırmalarında Yer Alan Her Bilim İnsanının İhtiyaç Duyduğu ve Kolayca Edinebileceği Beceri Seti
Review Article
ABSTRACTThe overwhelming advances in molecular modelling witnessed for a couple of decades go hand in hand with the booming computer and information technolo-gies. Computer-aided drug design (CADD) is probably the most important field of molecular modelling given the time scale and cost for turning a chemical entity into an approved drug. In this review we provide a brief definition of molecu-lar modelling and CADD with historical corner stones. In this review methods, tools, and applications of molecular modelling in different stages of CADD were focused on by referring to a number of success stories. Useful databases and non-commercial software for different purposes are also introduced. The review aims to provide a glimpse of these methods for scientists taking part in any field of drug research and to show that everyone can and should make the best of these methods with a vast amount of available free tools and documentation.
Keywords: molecular modelling, computer-aided drug design, virtual screening,
molecular docking, pharmacophore modelling, shape similarity, molecular dy-
namics simulations, ADMET prediction
ÖZETMoleküler modellemede son birkaç on yıldır tanıklık ettiğimiz baş döndürücü gelişmeler ilerleyen bilgisayar ve enformasyon teknolojileri ile birlikte gerçek-leşmektedir. Bilgisayar destekli ilaç tasarımı (BDİT), bir kimyasalın ilaca dönüş-türülmesi için gerekli zaman ve masraflar göz önüne alındığında, muhtemelen moleküler modellemenin en önemli alanıdır. Mevcut derlemede BDİT’nın farklı aşamalarındaki moleküler modelleme yöntemleri, araçları ve uygulamaları-na, bazı başarı hikayelerine de atıfta bulunularak odaklanılmıştır. Ayrıca, farklı amaçlar için kullanılan faydalı veri tabanları ve ticari olmayan yazılımlar tanıtıl-mıştır. Derleme, bu yöntemlerle ilgili, ilaç araştırmalarının herhangi bir alanında görev alan bilim insanlarına fikir vermeyi ve herkesin bu yöntemlerden, mevcut çok sayıda ücretsiz yazılım ve dokümantasyon ile en iyi şekilde faydalanabilece-ğini göstermeyi amaçlamaktadır.
Anahtar Kelimeler: moleküler modelleme; bilgisayar destekli ilaç tasarımı; sanal aktivite tarama; moleküler kenetleme; farmakofor modelleme; şekil benzerliği; moleküler dinamik simülasyonları; ADMET tahmini
Suat SARI*
Hacettepe University, Faculty of Pharmacy, Department of Pharmaceutical Chemistry, Ankara, Turkey
Corresponding author: Suat Sari Hacettepe University, Faculty of Pharmacy, Department of Pharmaceutical Chemistry, 06100 Sihhiye, Ankara,Turkey Tel.: + 90 312 305 18 72 Fax: + 90 312 305 32 72 E-mail: [email protected] , [email protected]
Received date : 18.04.2020 Accepted date : 26.05.2020
1. Introduction
The overwhelming advances in information technol-ogy observed in the past couple of decades, among many other things, made it possible to understand and predict molecules at every level of complexity by simply using our PCs or even our mobile de-vices. Thanks to multicore processors and high-end graphics available to almost everyone, molecular modelling is no longer the expertise of only those with highly technical computing skillset. Diversity of the methods in molecular modelling enabled this approach to reach to various fields in life sciences. Any researcher who crosses molecules’ path can get to know of many things in silico before or after an experiment (1, 2). Is my molecule water soluble enough to run an assay or how can I make it water soluble? Can this compound pass blood-brain bar-rier? Does it contain any functional groups that can react nonspecifically with the assay medium? Can it inhibit CYP3A4? Could this peptide assume an α-helix conformation? Which mutations could pos-sibly affect the function of my protein? Could this antagonist have triggered a desensitized state for my receptor? Questions like these can find more accu-rate answers in shorter time with the hands of the inquirers themselves thanks to molecular modelling.
Of course, molecular modelling has a special place in drug research. Given the time span and costs for a chemical entity to be labelled as “drug”, the ques-tions such as “which compound”, “which targets”, and “which off-targets” definitely need to be ad-dressed at the very beginning. At this point, molecu-lar modelling is put in use to work out what we call
“computer-aided drug design” (CADD) to save time and money, to dodge pit falls and dead ends, and to detect blind spots (3).
In this review, molecular modelling and virtual screening is defined and a historical background is provided with drug discovery perspective. Steps of CADD are presented with state-of-the-art methodol-ogies, applications, and success stories. The review also compiles resources for the most common tools of CADD.
2. Molecular Modelling and Its Historical Background
Molecular modelling is all the computational meth-ods used to predict molecular structure and behavior.
From material science to structural biology, molecu-lar modelling methodologies are used in many fields to understand systems made up of a wide range of complexity from small molecules to biological mac-romolecules such as proteins, receptors, and nu-cleotide chains. These systems can be modelled by treating atoms as particles with charges and potential energy using forcefields (molecular mechanics) or by applying wave function at atomic and subatomic scales (quantum mechanics) (4). From drawn repre-sentations of chemical structures to millisecond-long simulations of biological systems, molecular model-ling has undergone a massive progress through his-tory (5). Following are some of the milestones of molecular modelling:
The term chemical structure was introduced between 1858 and 1861 by identification of va-lence rules in organic chemistry and representing bonds as lines in molecules with carbon chains.
In 1865, August Wilhelm Hofmann for the first time used physical models in his organic chemis-try lecture, in which organic compounds such as methane and chloroform were represented with croquet balls joined by sticks.
Hofmann also established today’s commonly ac-cepted color scheme for atoms: black for carbon, white for hydrogen, blue for nitrogen and green for chlorine.
Crum-Brown in 1865 and Sir Edward Frankland and B. F. Duppa in 1867 were the scientists to use 2D drawn structures in ball and stick models for the first time.
That carbon compounds have tetrahedral geom-etry was first suggested by E. Paterno (1869), Jacobus Henricus van’t Hoff, and Joseph Achille LeBel (1874), which is considered as the emer-gence of three-dimensional molecular structure elaboration.
In 1898, van’t Hoff suggested that each carbon-carbon bond had a favored conformation about a torsional angle, setting the foundations of what we refer to as the global minimum energy confor-mations today.
Urey and Bradley introduced a formulation which included quadratic Hooke’s Law potential equa-tions to describe harmonic vibrations in simple molecules and found the Morse potential to give the best fit to empirical data for bond stretching
Hacettepe University Journal of the Faculty of Pharmacy
ISSN: 2458 - 880635
in 1931. This was a breakthrough in forcefield concept.
The first examples of molecular mechanics came in 1946 with Hill’s force field to treat the repulsive and attractive nonbonding terms, and quadratic terms for bond stretch and angle bend; Dostrovsky, Hughes and Ingold’s study on the repulsive and attractive nonbonding terms; and Westheimer and Mayer’s analyses of the confor-mations of hindered biphenyls. These methods let molecular modelling spread out of physical chemistry society.
The importance of 3D aspects of molecules in understanding structure, stability, conformation, and reactivity was appreciated with the emer-gence of conformation analysis led by the study of Barton on the conformations of steroids in 1950.
The most famous physical molecular model of all time, the structure of DNA double strand, was elucidated by Watson and Crick in 1953.
The first molecular dynamics (MD) calculations study was reported in the same year under the name Equation of State Calculations by Fast Computing Machines, which featured simulated annealing and aided the groundwork for Monte Carlo simulations.
Potentials of nonbonding interactions in organic structure modelling were first applied by Kitaig-orodsky in 1960.
The first use of computer for empirical force field calculations was reported in 1961 by Hendrick-son.
In 1965, Wiberg developed a steepest descent algorithm for geometry optimization to address conformational analysis.
Scientists from Oak Ridge National Laboratory in the US designed a molecular drawing program called ORTEP (Oak Ridge National Laboratory) in 1965. The US government also funded the first computer network in 1969, which is accepted as the ancestor of internet.
In 1971, Lee and Richards described the molecu-lar surface in protein structure context and eluci-dated an algorithm to derive it.
In 1974, computer modelling of oligosaccharides starting from crystal structure was reported for the first time. In the same year, force field cal-culations of synthetic macromolecules started to appear.
At the beginning of 1980s, with the booming per-sonal computer (PC) industry, molecular model-ling became accessible from PCs and the use of the graphical user interface (GUI) started. This marks the advent of personal molecular model-ling for the average chemist.
The World Wide Web kickstarted in 1993, which probably marks the beginning of web servers for molecular modelling.
3. Computer Aided Drug Design (CADD)
Drug design and discovery is highly complex and ex-pensive process that requires contribution of a wide range of disciplines. The general estimation is that it takes more than 10 years and a billion US dollars for a chemical entity to be used in the clinic. Although in silico methods are usually adopted in the early-to-mid-stage drug discovery studies, the selection of candidates passed from these stages to preclinical and clinical phases greatly affects the attrition rates. Therefore, the use of in silico methods in drug dis-covery has increased reasonably for the past couple
Figure 1. Steps of drug discovery and applications of CADD.
Hacettepe University Journal of the Faculty of Pharmacy
Volume 40 / Number 1 / January 2020 / pp. 34-47 Sari et al.36
of decades (Figure 1). CADD applications mainly assist experimental studies in decision making by following the major drug discovery stages, such as target validation, hit identification, lead generation, and optimization (3, 6).
3.1. Target Identification and Validation
Drug development process usually starts with identi-fication of a druggable target relevant to the disease of interest. Introduction of such targets prompts efforts to find potential compounds to modulate the target pathway and finally to elicit a phenotypic response (7). In silico approaches are usually helpful to un-derstand the structure, topology, and ligand binding sites of the target, as well as their key residues and ligand binding interactions. In silico methods such as homology modelling, molecular docking, and MD simulations have become paramount to assist in vitro studies for target validation such as protein structure elucidation (e.g. X ray diffraction), alanine scanning, site-directed mutagenesis, and radioligand binding (8-10).
Chemical and biological databases play crucial role in bringing the wet lab and computers together. Cur-rently, there is a vast number of databases which ar-chive and process countless sorts of data belonging to millions of molecules ranging from small mole-cules to proteins, genes, and nucleotide chains (Table 1) (11).
3.2. Hit Identification
Once a target is validated for a certain disorder, cam-paigns are run for identification of hit matter, i.e. chemical entities that display an intended activity against the target, by both pharma companies and academia. Just as the robotics technology made it possible to randomly screen biological activity of thousands of compounds in vitro at the same time (high-throughput screening), the computer technolo-gy did the same for screening millions of compounds in silico. With the so-called high-throughput virtual screening (or briefly virtual screening), biological screening of small compound sets rationally selected from huge libraries became more common. Virtual screening features a number of filters to narrow down the libraries step by step and to eventually sug-gest potential hits for in vitro tests (12). As the filters go from rough to exhaustive, the computational bur-den and time required increases dramatically, which, however is balanced with shrinking library through-
out the steps. Evaluation of druglike chemical space and PAINS (Pan-assay Interference Compounds) is an example for rough filters, which is commonly applied at the beginning of virtual screening cam-paigns (13, 14). This step is crucial for eliminat-ing entities with potentially poor pharmacokinetics and those with non-specific reactive functionalities, since such compounds account for a good amount of failed clinical candidates (15). This step is usually followed by a ligand- and/or structure-based virtual screening (16).
3.3. Lead Generation and Optimization
Finding hit compounds with ability to modulate an isolated target is the tip of the iceberg in drug dis-covery issues. When evaluated thoroughly (plotting dose-response curves, profiling for related targets, testing against cell cultures, in vivo and toxicity profiling, cytochrome P450 and efflux pump inter-actions, pharmacokinetics etc.) these hits will most likely fail at a certain point. This is where medici-nal chemists take the stage to tailor these hits to fit specific requirements (17). In silico drug discovery methods enable medicinal chemists to create librar-ies of countless virtual compounds envisaged accord-ingly. The so-called scaffold-hopping methodology utilizes both ligand- and structure-based approaches to model structurally relevant molecules as synthetic candidates (18). Establishing structure-activity rela-tionships (SARs) at this stage is crucial for creating virtual libraries. Quantitative SARs (QSAR) is an inevitable computation tool for decision making re-garding SARs (19). The synthesized compounds are then subjected to a set of in vitro and in vivo tests to obtain leads.
Lead optimization stage mainly focuses on pharma-cokinetic and toxicity issues, i.e. ADME+T (Absorp-tion, Distribution, Metabolism, Excretion + Toxic-ity), therefore in silico methods are less likely to be included at this level. However molecular modelling offers a wide range of tools to assist decision making for various scenarios. There are many web servers and free tools to evaluate ADME+T profile of com-pounds through fast in silico predictions, as well as more sophisticated, specific, and exhaustive models. Site of metabolism, hERG channel affinity, blood-brain barrier permeability, human serum albumin binding, CYP and P-glycoprotein affinity are among the properties that scientists can predict without a high level of expertise in computational chemistry
Hacettepe University Journal of the Faculty of Pharmacy
ISSN: 2458 - 880637
and bioinformatics (20-23).
4. Approaches in CADD
Rapid development in information technologies brought about a vast amount of data, which in-creased our ability to create highly predictive models. In CADD, data is related to either small molecules or their potential targets, i.e. biological macromol-ecules. Virtual screening approaches in CADD can be classified as ligand-based and structure/receptor-based methods according to the type of employed data. Ligand-based methods use models built exclu-
sively from ligand data, while structure-based or re-ceptor-based approaches benefit from structural data of target macromolecules. Virtual screening methods, in most cases, incorporate both approaches, which are referred to as hybrid methods. This, of course, has much to do with the extend of the data available (16).
4.1. Ligand-Based Methods
The general consideration in CADD is that similar li-gands are supposed to exert similar biological effects, which lies beneath the rationale of ligand-based
Table 1. Examples of useful chemical and biological databases Database 1 Source Description
ZINC https://zinc.docking.org Indexes more than 230M commercial compounds in ready-to-dock format with diverse data including predicted and biological properties and vendor information.
ChEMBL https://www.ebi.ac.uk/chembl/ Archives 2M compounds with detailed experimental data, target, assay, and literature information.
DrugBank www.drugbank.ca Includes 13,551 drug entries (approved and experimental) with detailed information including drug targets.
PubChem https://pubchem.ncbi.nlm.nih.gov Archives 103M compounds with relevant chemical, bioactivity, and literature information.
ChemSpider www.chemspider.com Provides 81M compounds with 278 data sources including supplier information.
DUD-E http://dude.docking.org Provides 22K active compounds with target information and 50 decoys for each active compound for virtual screening enrichment studies.
BindingDB www.bindingdb.org/bind Stores measured affinities of 820K small molecules to 7K protein targets.
Protein data banks
www.wwpdb.orgwww.pdbe.orgwww.rcsb.orgwww.bmrb.wisc.eduwww.pdbj.org
Store three-dimensional structural data of biological macromolecules. Currently, more than 160K structurers are available.
UniProt www.uniprot.org Provides sequence and functional information for some 177M proteins.
NCBI www.ncbi.nlm.nih.gov Along with literature (PubMed) and many other databases, NCBI is one of the largest protein, DNA, RNA, genome, and gene databases.
SCOP http://scop.mrc-lmb.cam.ac.uk Provides a detailed and comprehensive description of structural and evolutionary relationships between 532K proteins with known structure.
BioGRID https://thebiogrid.org Biological database of protein-protein interactions, genetic interactions, chemical interactions, and post-translational modifications.
PROSITE https://prosite.expasy.org Consists of documentation entries describing protein domains, families and functional sites as well as associated patterns and profiles to identify them.
1in alphabetical order
Hacettepe University Journal of the Faculty of Pharmacy
Volume 40 / Number 1 / January 2020 / pp. 34-47 Sari et al.38
CADD (24). Similarity can be measured in diverse ways, i.e. molecular descriptors, which incorporate molecular properties. According to the complexity and organization of the data, ligand-based methods are classified as 1D-, 2D- or 3D-methods (25). De-scriptors for 1D-methods include physicochemical properties of molecules without structural data, such as molecular weight, H bond acceptor count, and po-lar surface area. The idea of 1D-methods suggests that compounds belong to a specific group, such as orally available small-molecule drugs, possess similar set of physicochemical properties. This kind of methods are helpful to filter libraries in terms of drug-likeness or lead-likeness or to predict certain pharmacokinetic properties to help decision making (26). 2D-methods employ physicochemical proper-ties assigned to 2D-structure of compounds. Molecu-lar fingerprints are widely used to determine simi-larity between compounds regarding fragment con-nectivity independent from their spatial orientations, which is handled by 3D-methods, such as pharma-cophore modelling (Figure 2) and shape similarity (27-29). The increase in the order of dimensions causes computational burden. For example, in 3D-methods, similarity between two compounds is cal-culated by finding the best molecular volume align-ment, which is selected from possible conformations
of the compounds. This requires coordination data for each atom, which is re-calculated for each ge-ometry. However, increasing computational burden is usually the result of demands for higher precision in molecular modelling. In addition, 3D-geometry and spatial volume are valuable data in CADD con-sidering drug-target affinity. In an effort to reduce computation burden by maintaining precision, new methods were introduced in which 3D-molecular fields are reduced into 1D-descriptors or molecular volume information is compressed and described as vectors (30, 31).
4.2. Structure-Based Methods
Two things gave rise to the emergence of structure-based methods as powerful tools for CADD: advanc-es in biomolecular spectroscopic methods, such as X ray diffraction and NMR, and rapid development of computer technologies, especially processors and graphics. X ray diffractometers have become widely accessible, which triggered an avalanche of struc-tural data of biological macromolecules deposited in the web sites called protein data bank. Handling macromolecule data requires high-performing pro-cessors and graphics, which is available to individu-als, except clusters of parallel processors required in the case complex dynamics simulations. The impor-
Figure 2. Last-generation azole antifungals (A) and a pharmacophore model of the last-generation azole antifungals aligned with voriconazole (B). The model consists of three rings, two hydrophobic groups, and one H bond donor represented as brown rings, green spheres and a blue sphere with an arrow, respectively. The cut-off space for each pharmacophore is highlighted with a transparent sphere. Voriconazole is represented as gray sticks and balls.
Hacettepe University Journal of the Faculty of Pharmacy
ISSN: 2458 - 880639
tance of target macromolecule structure data became apparent with the ever expanding the knowledge of ligand-receptor affinity and interactions at molecu-lar level. CADD benefits from these developments in several situations some of which are listed below (32):
Creating a structural model for a protein when there is no spectroscopic or experimental struc-tural data
Identification of possible druggable sites for a re-ceptor
Ligand-receptor interactions and how these inter-actions manipulate signal transduction
Possible effects of certain residues, co-factors, ions, and solvents on conformational changes in macromolecule structure and its gating
The effects of charge polarization in ligand-re-ceptor binding
For these issues and more, there are powerful and popular tools, such as homology modelling, molecu-lar docking, and MD simulations, which are availa-ble either free of charge or through paid licenses (32).
4.2.1. Homology Modelling
Proteins with similar amino acid sequences tend to have similar tertiary structures. Homology modelling is a method to predict three-dimensional structure of a protein when there is no experimental structural data available. This is performed by the help of ho-mologous proteins with experimental structural data. The process starts with sequence alignment of the query structure with one or more template protein(s). This alignment and the atomic coordinates of the
template protein(s) are then used to thread the struc-tural model of the query protein. The model is further optimized by loop refinement and side chain mini-mization, and subjected to various structural assess-ment methods for validation (33, 34).
Apart from creating a whole structure model, homol-ogy modelling can be used to fill in loops and miss-ing side chains, something usually encountered with the structures obtained from protein data bank. Also, in silico modifications in protein structure such as alanine scanning is possible via homology model-ling. Thus, homology modelling is a powerful tool for structure-based modelling employed in target validation, hit generation, and optimization studies (35-37).
4.2.2. Molecular Docking
Molecular docking is an in silico technique to predict the preferred binding orientation and affinity of two molecules, a ligand and a receptor, to form a stable complex. Receptor in molecular docking is usually a macromolecule such as protein, DNA, RNA, or pep-tide, which is kept rigid; while ligand is a flexible small molecule. However, techniques for docking two macromolecules to each other or making both ligand and receptor flexible are available. Molecu-lar docking predicts countless ligand-receptor com-plexes (search space) and ranks them according to a score (docking score) which is calculated by a scor-ing function (Figure 3). Docking score is a metric used to predict how much affinity a ligand is bound to a receptor with, which is usually governed by non-bonded interactions, although it is possible to model possible covalent bonds between ligand and receptor by covalent docking method (38).
Figure 3. Molecular docking.
Hacettepe University Journal of the Faculty of Pharmacy
Volume 40 / Number 1 / January 2020 / pp. 34-47 Sari et al.40
Biological systems work by means of signal trans-duction, which is pretty much affected by the at-tributes of ligand-receptor complexes. By predicting ligand-receptor complexes it is possible to predict the outcomes of signal transduction, which makes molecular docking a precious tool for drug design. In recent years, molecular docking has become an indispensable component of virtual screening. On the other hand, it is routinely applied to study ligand-receptor interactions at atomic level to understand the importance of certain amino acids, cofactors, chelating with metals, hydrogen and halogen bonds, solvation effects, and more (Figure 4) (39).
4.2.3. MD Simulations
While molecular docking provides a picture of a biological process, it is the dynamics of this process that actually matters for its biological consequences. MD of a system is the physical movements of all its atoms, which is simulated by adding Newtonian me-chanics to the initial conditions of each atom, i.e. en-ergies and coordinates. Then, forces between atoms or particles and potential energies are calculated at each given time period to determine trajectories for atoms or molecules, which reflect the dynamic evo-lution of the system (42).
Depending on the number of the atoms of the sys-tem and the presumed time scale, MD simulations
require a great amount of data usage and time, espe-cially compared to molecular docking or other target and ligand-based methods. To overcome this com-putational burden in a reasonable amount of time, parallel processors were introduced, some of which are available to researchers via remote access. Lately, an evolutionary method called Graphics Processing Unit (GPU) acceleration has made use of high-end graphics processors to greatly reduce computation time, making researchers less dependent to costly CPU clusters (43).
In addition to target validation, MD simulations are gradually becoming a part of virtual screening cam-paigns thanks to the advances mentioned above. MD simulations are usually considered in connection with molecular docking in virtual screening, for ex-ample to determine the stability of a ligand-receptor complex, however its connected use with ligand-based methods is becoming increasingly popular (44, 45).
5. CADD: Success Stories
In contrast with what one would expect, most of the chemical entities labelled as “drug” today come from serendipitous studies such as combinatorial chemistry (46). This is partly due to a sharp decrease in the speed of new entities hitting the market ob-served past couple of decades, when CADD is most
Figure 4. Superposition of the co-crystallized ligand (oteseconazole) and its docking pose obtained by Glide (2019-4, Schrödinger, LLC, NY, USA) (40) in Candida albicans lanosterol 14α demethylase (CACYP51) active site (A) and binding interactions of oteseconazole with CACYP51 according to the crystallographic study (B) (41). The images were rendered using Maestro (2019-4, Schrödinger, LLC, NY, USA).
Hacettepe University Journal of the Faculty of Pharmacy
ISSN: 2458 - 880641
expected to bear fruit. Still, examples of drugs dis-covered through CADD inspire researchers from dif-ferent backgrounds, three of which are provided as example in this review.
Dorzolamide (Merck): Marketed as an eye-drop for glaucoma, dorzolamide was developed through structure-based drug design (47). This compound was developed as a human carbonic anhydrase II (hCAII) inhibitor and tailored to fit in the specific active site of hCAII starting from a parent racemate compound (MK-927), whose enantiomers show 100-fold difference in affinity to hCAII (48, 49). X ray crystallography studies and ab initio calculations identified the conformational variations that caused the observed potency difference between the enan-tiomers. Therefore, further modifications were taken on for the S enantiomer of MK-927 using structure-based modelling and introduction of a methyl group to the 6th position yielded a more potent derivative. The 4-isobuthylamino group was replaced with eth-ylamino group to counter the reduced water solubil-ity. The resulting four enantiomers were evaluated through in vitro and crystallographic studies and the trans S,S configuration (dorzolamide) was found to have the greatest potency (50).
Zanamivir (GSK) and Oseltamivir (Gilead Sci-ences): The antiviral drug zanamivir acts through viral neuraminidase inhibition and is used against influenza infections (51). Following the structural elucidation of neuraminidase via X ray crystallog-raphy, structure-based virtual screening campaigns were conducted to find potential anti-viral inhibitors and zanamivir is the result of one of these campaigns using the software GRID (52). Zanamivir was de-signed from 2-deoxy-2,3-dehydro-N-acetylneu-raminic acid by substitution of the 4-hydroxy group with 4-guanidino. Further structure-based design to utilize an extra binding gorge on neuraminidase led to the discovery of more potent and orally available oseltamivir (47).
Aliskiren (Novartis): Aliskiren was designed as a renin inhibitor for the treatment of hypertension. It is also the first member of the class called direct renin inhibitors. Aliskiren’s design starts with the efforts to find renin inhibitors by mimicking the natural peptide substrate of the renin system (53). The first non-peptide derivative was developed by Goschke et al. (1997), which was a success compared to the previous peptide derivatives in terms of pharmacoki-
netic profile (54). With a structure-based modelling using a crystal structure of renin, the lead designed by Goschke et al. was further optimized to improve potency. Identification of an additional pocket in the renin catalytic site through the following X ray crystallography studies led to rational design of new derivatives with better affinity and selectivity to re-nin over other aspartic peptidases and further SAR studies to optimize in silico interactions with renin resulted in a derivative, which was aliskiren, with potency at sub-nanomolar concentration (55).
6. Resources for Molecular Modelling Tools
Thanks to the age of information and the growing “open source, open data” trend, it has never been easier to understand and utilize molecular modelling tools of different sorts. The internet is full of free-to-use applications or web servers as well as tutorial materials (Table 2).
Increasing popularity of molecular modelling cre-ated demands for hands-on training and workshops, which are now a common part of scientific meetings of related fields. These workshops, some of which are supported by the leading molecular modelling software companies, offer important opportunities, especially for postgraduates and postdoctoral fel-lows. In this respect, we organized the 1st Molecular Modelling Workshop in Hacettepe University Fac-ulty of Pharmacy with Hacettepe University Me-dicinal Chemistry Research, Development, and Ap-plication Center (MAGUM) on December 5-6, 2019. The workshop focused on state-of-the-art techniques in molecular modelling for CADD (56). The par-ticipants had the opportunity to know academic free modelling tools and experience hands-on training. The workshop also provided a general idea, opportu-nities, catches, and pitfalls of virtual screening.
7. Conclusions
The overwhelming developments in computer and information technologies for the past couple of dec-ades have boosted molecular modelling and its key application, CADD. As computers became an indis-pensable material of research, so did molecular mod-elling techniques in CADD at all levels, from sketch-ing a molecule to running millisecond MD simula-
Hacettepe University Journal of the Faculty of Pharmacy
Volume 40 / Number 1 / January 2020 / pp. 34-47 Sari et al.42
Tabl
e 2.
Lis
t of e
xam
ple
non-
com
mer
cial
tool
s for
mol
ecul
ar m
odel
ling
and
CA
DD
1 in
alph
abet
ical
ord
er
Soft
war
e 1
Type
App
licat
ion
Sour
ceL
icen
seD
evel
oper
(s)
Aut
oDoc
kSt
and-
alon
e so
ftwar
eM
olec
ular
doc
king
http
://au
todo
ck.sc
ripps
.edu
Free
Th
e Sc
ripps
Res
earc
h In
stitu
te
Aut
oDoc
k V
ina
Stan
d-al
one
softw
are
Mol
ecul
ar d
ocki
nght
tp://
vina
.scrip
ps.e
duFr
ee
The
Scrip
ps R
esea
rch
Inst
itute
Avog
adro
Stan
d-al
one
softw
are
3D m
olec
ular
edi
tion
and
mol
ecul
ar
mec
hani
cs c
alcu
latio
ns
http
s://a
voga
dro.
ccFr
ee
Uni
vers
ity o
f Pitt
sbur
gh
DO
CK
Stan
d-al
one
softw
are
Mol
ecul
ar d
ocki
nght
tp://
dock
.com
pbio
.ucs
f.edu
Aca
dem
ic fr
eeU
nive
rsity
of C
alifo
rnia
-Sa
n Fr
anci
sco
Chi
mer
aSt
and-
alon
e so
ftwar
eIn
tera
ctiv
e st
ruct
ure
visu
aliz
atio
nht
tps:
//ww
w.cg
l.ucs
f.edu
/chi
mer
aFr
eeU
nive
rsity
of C
alifo
rnia
-Sa
n Fr
anci
sco
DS
Vis
ualiz
erSt
and-
alon
e so
ftwar
eIn
tera
ctiv
e st
ruct
ure
visu
aliz
atio
nht
tps:
//ww
w.3d
sbio
via.
com
/pro
duct
s/co
llabo
rativ
e-sc
ienc
e/bi
ovia
-dis
cove
ry-
stud
io/v
isua
lizat
ion-
dow
nloa
d.ph
pA
cade
mic
free
BIO
VIA
GR
OM
AC
SSt
and-
alon
e so
ftwar
eM
D si
mul
atio
nsht
tp://
ww
w.gr
omac
s.org
Free
Sc
ienc
e fo
r Life
La
bora
tory
, Sto
ckho
lm
Uni
vers
ity
HA
DD
OC
KSt
and-
alon
e so
ftwar
ePr
otei
n-pr
otei
n do
ckin
ght
tp://
milo
u.sc
ienc
e.uu
.nl/s
ervi
ces/
HA
DD
OC
K2.
2A
cade
mic
free
Cen
tre B
ijvoe
t Cen
ter f
or
Bio
mol
ecul
ar R
esea
rch
I-TA
SSER
Web
serv
erC
ompa
rativ
e pr
otei
n m
odel
ling
http
s://z
hang
lab.
ccm
b.m
ed.u
mic
h.ed
u/I-
TASS
ERFr
ee
Zhan
g La
b, U
nive
rsity
of
Mic
higa
n
KN
IME
Stan
d-al
one
softw
are
Des
crip
tor c
alcu
latio
n,
QSA
R m
odel
ling
ww
w.k
nim
e.co
mFr
ee
KN
IME
AG
Mae
stro
Stan
d-al
one
softw
are
Inte
ract
ive
stru
ctur
e vi
sual
izat
ion
http
s://w
ww.
schr
odin
ger.c
om/fr
eem
aest
roA
cade
mic
free
Schr
ödin
ger
Mar
vinS
ketc
hSt
and-
alon
e so
ftwar
eSm
all m
olec
ule
sket
ch
and
editi
onht
tps:
//che
max
on.c
om/p
rodu
cts/
mar
vin
Aca
dem
ic fr
eeC
hem
Axo
n
Mem
oir
Web
serv
erM
embr
ane
prot
ein
mod
ellin
ght
tp://
opig
.stat
s.ox.
ac.u
k/w
ebap
ps/m
emoi
rFr
ee
Oxf
ord
Prot
ein
Info
rmat
ics
Gro
up, O
xfor
d U
nive
rsity
Mod
elle
rSt
and-
alon
e so
ftwar
eC
ompa
rativ
e pr
otei
n m
odel
ling
http
s://s
alila
b.or
g/m
odel
ler
Free
U
nive
rsity
of C
alifo
rnia
Sa
n Fr
anci
sco
Mol
insp
iratio
nW
eb se
rver
Des
crip
tor c
alcu
latio
nw
ww
.mol
insp
iratio
n.co
mFr
ee
Mol
insp
iratio
n C
hem
info
rmat
ics
Mol
Prob
itySt
and-
alon
e so
ftwar
e
Fixi
ng a
nd h
ydro
gen
mod
ellin
g fo
r PD
B
stru
ctur
esht
tp://
mol
prob
ity.b
ioch
em.d
uke.
edu
Free
Sc
hool
of M
edic
ine,
Duk
e U
nive
rsity
Hacettepe University Journal of the Faculty of Pharmacy
ISSN: 2458 - 880643
NA
MD
and
VM
DSt
and-
alon
e so
ftwar
eM
D si
mul
atio
nsht
tps:
//ww
w.ks
.uiu
c.ed
u/D
evel
opm
ent
Aca
dem
ic fr
eeTh
eore
tical
and
C
ompu
tatio
nal B
ioph
ysic
s G
roup
Ope
nEye
Stan
d-al
one
softw
are
Mol
ecul
ar m
odel
ling
and
CA
DD
pla
tform
ww
w.e
yeso
pen.
com
Ope
nEye
Sci
entifi
c
Phar
maG
ist
Web
serv
erPh
arm
acop
hore
m
odel
ling
http
://bi
oinf
o3d.
cs.ta
u.ac
.il/P
harm
aGis
tFr
ee
Scho
ol o
f Com
pute
r Sc
ienc
e, T
el A
viv
Uni
vers
ity
Phar
mer
Stan
d-al
one
softw
are
Phar
mac
opho
re
mod
ellin
ght
tp://
smoo
thdo
ck.c
cbb.
pitt.
edu/
phar
mer
Free
C
amac
ho L
ab, U
nive
rsity
of
Pitt
sbur
gh
Phar
mM
appe
rW
eb se
rver
Phar
mac
opho
re
mod
ellin
gw
ww
.lila
b-ec
ust.c
n/ph
arm
map
per
Free
H
ongl
in L
i’s L
ab, E
ast
Chi
na U
nive
rsity
of
Scie
nce
& T
echn
olog
y
PPM
Ser
ver
Web
serv
erPr
edic
tion
of m
embr
ane
prot
ein
orie
ntat
ion
http
s://o
pm.p
har.u
mic
h.ed
u/pp
m_s
erve
rFr
eeLo
miz
e G
roup
, Uni
vers
ity
of M
ichi
gan
PRO
CH
ECK
Web
serv
erPr
otei
n st
ruct
ure
valid
atio
nht
tps:
//ser
vice
sn.m
bi.u
cla.
edu/
PRO
CH
ECK
Free
Uni
vers
ity o
f Cal
iforn
ia-
San
Fran
cisc
o
QSA
R T
oolb
oxSt
and-
alon
e so
ftwar
eQ
SAR
mod
ellin
ght
tps:
//qsa
rtool
box.
org
Free
The
Labo
rato
ry o
f M
athe
mat
ical
Che
mis
try,
Prof
. Dr.
As.
Zlat
arov
U
nive
rsity
QSA
RIN
S-C
hem
Stan
d-al
one
softw
are
QSA
R m
odel
ling
http
://w
ww.
qsar
.itFr
ee
The
QSA
R R
esea
rch
Uni
t in
Envi
ronm
enta
l C
hem
istry
and
Ec
otox
icol
ogy,
Uni
vers
ity
of In
subr
ia
Ros
etta
Stan
d-al
one
softw
are
De
novo
pro
tein
m
odel
ling
http
s://w
ww.
rose
ttaco
mm
ons.o
rg/s
oftw
are
Aca
dem
ic fr
eeR
oset
taC
omm
ons
Silic
os-it
Col
lect
ion
of
stan
d-al
one
softw
are
Liga
nd-b
ased
virt
ual
scre
enin
ght
tp://
silic
os-it
.be.
s3-w
ebsi
te-e
u-w
est-1
.am
azon
aws.c
omFr
ee
Silic
os-it
Swis
sAD
ME
Web
serv
erD
escr
ipto
r cal
cula
tion,
A
DM
ET p
redi
ctio
nw
ww
.swis
sadm
e.ch
Free
Sw
iss I
nstit
ute
of
Bio
info
rmat
ics
Swis
sDoc
kW
eb se
rver
Mol
ecul
ar D
ocki
ngw
ww
.swis
sdoc
k.ch
Free
Sw
iss I
nstit
ute
of
Bio
info
rmat
ics
SWIS
S-M
OD
ELW
eb se
rver
Com
para
tive
prot
ein
mod
ellin
ght
tps:
//sw
issm
odel
.exp
asy.
org
Free
Sw
iss I
nstit
ute
of
Bio
info
rmat
ics
Swis
sSim
ilarit
yW
eb se
rver
Liga
nd-b
ased
virt
ual
scre
enin
gw
ww
.swis
ssim
ilarit
y.ch
Free
Sw
iss I
nstit
ute
of
Bio
info
rmat
ics
Tabl
e 2.
con
tinue
d...
Hacettepe University Journal of the Faculty of Pharmacy
Volume 40 / Number 1 / January 2020 / pp. 34-47 Sari et al.44
tions, for scientists of drug research. There are many examples, in which the cost and time needed for drug discovery is reduced, complex decision-making pro-cesses are eased, disease pathways and interplay of diverse molecules are understood. In addition to availability of fast-processing computers, exponen-tially increasing amount of open data in this field is encouraging to learn and utilize molecular modelling in CADD. Actually, today there are more molecular modelling tools that are free or academic free than commercial software with paid license. The ocean called “internet” is full of documentations, tutori-als, and forums to help through getting familiar with these tools and troubleshooting. Also, molecular modelling workshops, separate or as part of scientif-ic meetings, became very common and wide-spread. Thus, every scientist in drug research can and should make the best of molecular modelling.
References
1. Dalkas GA, Vlachakis D, Tsagkrasoulis D, Kastania A, Kos-sida S: State-of-the-art technology in modern computer-aided drug design. Briefings in Bioinformatics 2013, 14(6):745-752.
2. Ooms F: Molecular modeling and computer aided drug design. Examples of their applications in medicinal chemistry. Cur-rent Medicinal Chemistry 2000, 7(2):141-58.
3. Kapetanovic IM: Computer-aided drug discovery and deve-lopment (CADDD): in silico-chemico-biological approach. Chemico-Biological Interactions 2008, 171(2):165-176.
4. Leach AR: Molecular Modelling : Principles and Applications. Pearson Education Ltd.; Essex, UK, 2001.
5. Schlecht MF: Molecular Modeling on the PC. Wiley-VCH; New York, USA, 1997.
6. Ramirez D: Computational Methods Applied to Rational Drug Design. The Open Medicinal Chemistry Journal 2016, 10:7-20.
7. Blake RA: Target validation in drug discovery. Methods in Molecular Biology 2007, 356:367-77.
8. Agamah FE, Mazandu GK, Hassan R, Bope CD, Thomford NE, Ghansah A, et al.: Computational/in silico methods in drug target and lead prediction. Briefings in Bioinformatics 2019, bbz103.
9. Sahu TK, Pradhan D, Rao AR, Jena L: In silico site-directed mutagenesis of neutralizing mAb 4C4 and analysis of its inte-raction with G-H loop of VP1 to explore its therapeutic app-lications against FMD. Journal of Biomolecular Structure and Dynamics 2019, 37(10):2641-2651.
10. Anand P, Nagarajan D, Mukherjee S, Chandra N: ABS-Scan: In silico alanine scanning mutagenesis for binding site residu-es in protein-ligand complex. F1000Research 2014, 3:214.
11. Audouze K, Taboureau O: Chemical biology databases: from aggregation, curation to representation. Drug Discovery To-day: Technologies 2015, 14:25-29.
12. Hughes JP, Rees S, Kalindjian SB, Philpott KL: Principles of early drug discovery. British Journal of Pharmacology 2011, 162(6):1239-1249.
13. Schneider N, Jackels C, Andres C, Hutter MC: Gradual in silico filtering for druglike substances. Journal of Chemical Information and Modeling 2008, 48(3):613-628.
14. Baell J, Walters MA: Chemistry: Chemical con artists foil drug discovery. Nature 2014, 513(7519):481-483.
15. Hay M, Thomas DW, Craighead JL, Economides C, Rosenthal J: Clinical development success rates for investigational drugs. Nature Biotechnology 2014, 32(1):40-51.
16. Katsila T, Spyroulias GA, Patrinos GP, Matsoukas MT: Com-putational approaches in target identification and drug disco-very. Computational and Structural Biotechnology Journal 2016, 14:177-184.
17. Keseru GM, Makara GM: Hit discovery and hit-to-lead appro-aches. Drug Discovery Today 2006, 11(15-16):741-748.
18. Zhao H. Scaffold selection and scaffold hopping in lead ge-neration: a medicinal chemistry perspective. Drug Discovery Today 2007, 12(3-4):149-155.
19. Subramaniyan B: QSAR and Lead Optimization. In: Puratchi-kody A, Lakshmana Prabu S, Umamaheswari A (eds.), Com-puter Applications in Drug Discovery and Development, IGI Global; Pennsylvania, USA. 2018: pp 21.
20. Hosea NA, Jones HM: Predicting pharmacokinetic profiles using in silico derived parameters. Molecular Pharmaceutics 2013, 10(4):1207-1215.
21. Daina A, Michielin O, Zoete V: SwissADME: a free web tool to evaluate pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small molecules. Scientific Reports 2017, 7:42717.
22. Chemi G, Gemma S, Campiani G, Brogi S, Butini S, Brindisi M: Computational Tool for Fast in silico Evaluation of hERG K(+) Channel Affinity. Frontiers in Chemistry 2017, 5:7.
23. Bikadi Z, Hazai I, Malik D, Jemnitz K, Veres Z, Hari P, et al.: Predicting P-glycoprotein-mediated drug transport based on support vector machine and three-dimensional crystal structu-re of P-glycoprotein. PLoS One 2011, 6(10):e25815.
24. Martin YC, Kofron JL, Traphagen LM: Do structurally similar molecules have similar biological activity? Journal of Medici-nal Chemistry 2002, 45(19):4350-4358.
Hacettepe University Journal of the Faculty of Pharmacy
ISSN: 2458 - 880645
25. Zoete V, Daina A, Bovigny C, Michielin O. SwissSimilarity: A Web Tool for Low to Ultra High Throughput Ligand-Based Virtual Screening. Journal of Chemical Information and Mo-deling 2016, 56(8):1399-1404.
26. Bajorath J: Selected concepts and investigations in compound classification, molecular descriptor analysis, and virtual scre-ening. Journal of Chemical Information and Modeling 2001, 41(2):233-245.
27. Cereto-Massague A, Ojeda MJ, Valls C, Mulero M, Garcia-Vallve S, Pujadas G: Molecular fingerprint similarity search in virtual screening. Methods 2015, 71:58-63.
28. Caporuscio F, Tafi A: Pharmacophore modelling: a forty year old approach and its modern synergies. Current Medicinal Chemistry 2011, 18(17):2543-2553.
29. Kumar A, Zhang KYJ: Advances in the Development of Shape Similarity Methods and Their Application in Drug Discovery. Frontiers in Chemistry 2018, 6:315.
30. Thijs G, Langenaeker W, De Winter H: Application of spectrophores™ to map vendor chemical space using self-organising maps. Journal of Cheminformatics 2011, 3(1):P7.
31. Ballester PJ, Richards WG: Ultrafast shape recognition to se-arch compound databases for similar molecular shapes. Jour-nal of Computational Chemistry 2007, 28(10):1711-1723.
32. Batool M, Ahmad B, Choi S: A Structure-Based Drug Disco-very Paradigm. International Journal of Molecular Sciences 2019, 20(11): E2783.
33. Muhammed MT, Aki-Yalcin E: Homology modeling in drug discovery: Overview, current applications, and future perspec-tives. Chemical Biology & Drug Design 2019, 93(1):12-20.
34. Schwede T: Homology Modeling of Protein Structures. In: Roberts GCK (eds), Encyclopedia of Biophysics. Springer; Berlin, Germany. 2013: pp 992-998.
35. Fiser A, Sali A: ModLoop: automated modeling of loops in protein structures. Bioinformatics 2003, 19(18):2500-2501.
36. Liang S, Grishin NV: Side-chain modeling with an optimized scoring function. Protein Science 2002, 11(2):322-331.
37. Saranyah K, Kalva S, Mukund N, Singh SK, Saleena LM: Ho-mology modeling and in silico site directed mutagenesis of pyruvate ferredoxin oxidoreductase from Clostridium thermo-cellum. Combinatorial Chemistry & High Throughput Scree-ning 2015, 18(10):975-989.
38. Meng XY, Zhang HX, Mezei M, Cui M: Molecular docking: a powerful approach for structure-based drug discovery. Cur-rent Computer-Aided Drug Design 2011, 7(2):146-157.
39. de Ruyck J, Brysbaert G, Blossey R, Lensink MF: Molecular docking as a popular tool in drug design, an in silico travel. Advances and Applications in Bioinformatics and Chemistry 2016, 9:1-11.
40. Friesner RA, Banks JL, Murphy RB, Halgren TA, Klicic JJ, Mainz DT, et al.: Glide: a new approach for rapid, accurate docking and scoring. 1. Method and assessment of docking accuracy. Journal of Medicinal Chemistry 2004, 47(7):1739-1749.
41. Hargrove TY, Friggeri L, Wawrzak Z, Qi A, Hoekstra WJ, Schotzinger RJ, et al.: Structural analyses of Candida albicans sterol 14alpha-demethylase complexed with azole drugs add-ress the molecular basis of azole-mediated inhibition of fun-gal sterol biosynthesis. Journal of Biological Chemistry 2017, 292(16):6728-6743.
42. Haile JM: Molecular Dynamics Simulation: Elementary Met-hods. Wiley; Chichester, UK, 1992.
43. Friedrichs MS, Eastman P, Vaidyanathan V, Houston M, Leg-rand S, Beberg AL, et al.: Accelerating molecular dynamic simulation on graphics processing units. ournal of Computati-onal Chemistry 2009, 30(6):864-872.
44. Salmaso V, Moro S: Bridging Molecular Docking to Molecu-lar Dynamics in Exploring Ligand-Protein Recognition Pro-cess: An Overview. Frontiers in Pharmacology 2018, 9:923.
45. Wieder M, Perricone U, Seidel T, Langer T: Pharmacophore Models Derived From Molecular Dynamics Simulations of Protein-Ligand Complexes: A Case Study. Natural Product Communications 2016, 11(10):1499-1504.
46. Ban TA: The role of serendipity in drug discovery. Dialogues in Clinical Neuroscience 2006, 8(3):335-344.
47. Talele TT, Khedkar SA, Rigby AC: Successful applications of computer aided drug discovery: moving drugs from concept to the clinic. Current Topics in Medicinal Chemistry 2010, 10(1):127-141.
48. Eriksson AE, Jones TA, Liljas A: Refined structure of hu-man carbonic anhydrase II at 2.0 A resolution. Proteins 1988, 4(4):274-282.
49. Baldwin JJ, Ponticello GS, Anderson PS, Christy ME, Murc-ko MA, Randall WC, et al.: Thienothiopyran-2-sulfonamides: novel topically active carbonic anhydrase inhibitors for the treatment of glaucoma. Journal of Medicinal Chemistry 1989, 32(12):2510-2513.
50. Greer J, Erickson JW, Baldwin JJ, Varney MD: Application of the three-dimensional structures of protein target molecules in structure-based drug design. Journal of Medicinal Chemistry 1994, 37(8):1035-1054.
51. McKimm-Breschkin JL: Influenza neuraminidase inhibitors: antiviral action and mechanisms of resistance. Influenza and Other Respiratory Viruses 2013, 7 Suppl 1:25-36.
52. von Itzstein M, Wu WY, Kok GB, Pegg MS, Dyason JC, Jin B, et al.: Rational design of potent sialidase-based inhibitors of influenza virus replication. Nature 1993, 363(6428):418-423.
Hacettepe University Journal of the Faculty of Pharmacy
Volume 40 / Number 1 / January 2020 / pp. 34-47 Sari et al.46
53. Wolfenden R: Analog approaches to the structure of the tran-sition state in enzyme reactions. Accounts of Chemical Rese-arch 1972, 5(1):10-18.
54. Göschke R, Cohen NC, Wood JM, Maibaum J: Design and synthesis of novel 2,7-dialkyl substituted 5(S)-amino-4(S)-hydroxy-8-phenyl-octanecarboxamides as in vitro potent pep-tidomimetic inhibitors of human renin. Bioorganic & Medici-nal Chemistry Letters 1997, 7(21):2735-2740.
55. Rahuel J, Rasetti V, Maibaum J, Rueger H, Goschke R, Cohen NC, et al.: Structure-based drug design: the discovery of novel nonpeptide orally active inhibitors of human renin. Chemical Biology 2000, 7(7):493-504.
56. Sarı S. Birinci Moleküler Modelleme Çalıştayı. Hacettepe Üniversitesi Eczacılık Fakültesi Medi̇si̇nal Ki̇mya Araştırma, Geli̇şti̇rme ve Uygulama Merkezi̇. 5-6 Aralık 2019, Ankara, Türkiye.
Hacettepe University Journal of the Faculty of Pharmacy
ISSN: 2458 - 880647